On 11/14/2013 10:36 AM, santhu kumar wrote: > Hi Gabriel and Chales, > > Thanks for the response. > > @Gabriel : The percent deviation seems like a good indicator for most > of the RDC values except those very small RDC values. I am not sure if > thats the right metric over-all. In almost all the RDC's I am using, > there seems to be 1-2 RDC's of this small magnitude. > it depends on the molecule and alignment tensor, some proteins have dozens very small RDCs (e.g. if a helix is oriented at magic angle relative to the longest alignment tesnor axis)
> @Charles : Can you please elaborate a bit more about semi-quantitative > errors, |err| .. Is that the experimental observation errors ? or the > RMS between Experimental and calculated RDC's. [err] must be experimental errors, they are generally inversely proportional to the S/N > > And also I have questions about modelling errors in RDC's. In the work > you have did, you used 0.26 HZ as the standard deviation for NH for > synthetic data, but when you have experimental data of magnitude 0.03 > [less than the standard deviation used in error modelling], is this > approach still valid? > > Thanks > santhosh > > > On Thu, Nov 14, 2013 at 10:00 AM, Charles Schwieters > <charles at schwieters.org <mailto:charles at schwieters.org>> wrote: > > > Hello Santhosh-- > > > > > To find how good a refinement is, R-factors or Q-factor is used > as a general > > broad indicator. > > > > R-factor defined as : sqrt{Sum[(Exp-obs)^2])/Sum[Exp^2]/2} > > > > This R-factor is a good indicator of the over-all data fit but > is there any such > > indicator for residue wise improvement. I tried residue wise > R-factor as : > > > > Residue R-factor : |Exp - Obs| / |Exp| > > [basically percent deviation]. > > > > Gabriel is correct that R-factor does take into account the magnitude > of RDCs, but for a per-residue metric, if you have semi-quantitative > errors, how about this residual: > > |Exp - Calcd| / |err| > > ?? > > Charles > > > > > _______________________________________________ > Xplor-nih mailing list > Xplor-nih at cake.cit.nih.gov > http://cake.cit.nih.gov/mailman/listinfo/xplor-nih -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Gabriel Cornilescu, Ph.D. Associate Researcher, NMRFAM University of Wisconsin - Madison Biochemistry Department 433 Babcock Drive, Rm. B160C Madison, WI 53706-1544 Google Phone: (608)-301-5876 Fax: (608)-262-3759 E-mail: gabrielc at nmrfam.wisc.edu http://www.nmrfam.wisc.edu/ https://mywebspace.wisc.edu/gcornilescu/web/ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://cake.cit.nih.gov/pipermail/xplor-nih/attachments/20131114/e2b44e2f/attachment-0001.html>
